Kimi K4 and GLM 5.5 Flash records appear on OpenCode: What is confirmed?

Public OpenCode data pages now carry records named Kimi K4, GLM 5.4, GLM 5.5 Flash, DeepSeek V4.1 Pro and Muse Spark 1.4 Contributor. A check at 13:25 Turkey time on September 27, 2026 confirmed that the pages exist. Their existence alone does not establish that the models have been announced or can be used through OpenCode.

What do the pages show?

OpenCode’s pages for Kimi K4, GLM 5.4, GLM 5.5 Flash, DeepSeek V4.1 Pro and Muse Spark 1.4 Contributor share an important distinction: Some show completed sessions, yet all five show zero tokens and zero users. Context window, maximum output and release date are marked unknown.

There is not enough evidence to interpret a “completed session” as a successful model call or a useful test. In particular, the zero-token and zero-user fields do not explain how the session counts arose. The pages also provide no basis for claims about performance, price or capabilities.

Can developers use these models?

At the time of the check, none of these five exact model IDs appeared in OpenCode’s Zen model list. Other names in that list, including qwen3.8-max, do not establish access to the five records discussed here. A data entry and a selectable API model are different things. Developers planning an integration should check the exact model ID in the list and verify access through their own account before relying on it.

This check captures one moment. If the list changes later, the new state needs a fresh assessment; absence today cannot rule out future access. Equally, opening a page by URL does not show that a model has an endpoint accepting requests. An application decision needs more than a name: verified access, documented limits and the result of an actual call. These records provide none of those checks, nor do they contain a published model card or provider announcement.

One OpenCode comparison page also displays 50/100 for GLM 5.5 Flash. That figure should not be treated as a benchmark: The page explicitly labels it “No data,” making it a neutral placeholder where measurements are unavailable. It cannot support a quality ranking against other models.

What remains unknown?

The names appear under record paths associated with Moonshot, Zhipu, DeepSeek and Meta; they do not amount to an announcement of OpenCode’s own model weights. The official pages reviewed offered no confirmation that the records were exposed by accident or that releases are imminent. Context and output limits, release dates, prices and actual availability remain unknown. Until an announcement or a clear change in the Zen list provides more evidence, the confirmed story is the appearance of the names on data pages, with the counters shown above.